U.S. patent number 11,061,382 [Application Number 16/223,131] was granted by the patent office on 2021-07-13 for methods of forming electroformed components and related system.
This patent grant is currently assigned to General Electric Company. The grantee listed for this patent is GENERAL ELECTRIC COMPANY. Invention is credited to Andrew Joseph Detor, Pei-Hsin Kuo.
United States Patent |
11,061,382 |
Detor , et al. |
July 13, 2021 |
Methods of forming electroformed components and related system
Abstract
A method of forming a component by an electroforming process
using an electroforming apparatus is presented. The electroforming
apparatus includes an anode, a cathode and an electrolyte including
a metal salt. The method includes receiving a set of training
electroforming process parameters; training a machine learning
algorithm based on at least a subset of the set of training
electroforming process parameters; generating a set of updated
operating electroforming parameters from the trained machine
learning algorithm; and operating the electroforming apparatus
based on the set of updated operating electroforming parameters.
The step of operating the electroforming apparatus includes
applying an electric current between the anode and the cathode in
the presence of the electrolyte and depositing a plurality of metal
layers on a cathode surface to form the component. A system of
forming a component is also presented.
Inventors: |
Detor; Andrew Joseph (Burnt
Hills, NY), Kuo; Pei-Hsin (Alplaus, NY) |
Applicant: |
Name |
City |
State |
Country |
Type |
GENERAL ELECTRIC COMPANY |
Schenectady |
NY |
US |
|
|
Assignee: |
General Electric Company
(Schenectady, NY)
|
Family
ID: |
1000005673501 |
Appl.
No.: |
16/223,131 |
Filed: |
December 18, 2018 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20200192328 A1 |
Jun 18, 2020 |
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
C25D
21/12 (20130101); G05B 19/4083 (20130101); G06N
20/00 (20190101); C25D 1/003 (20130101); G05B
2219/49007 (20130101); G05B 2219/33034 (20130101) |
Current International
Class: |
C25D
21/12 (20060101); G05B 19/408 (20060101); G06N
20/00 (20190101); C25D 1/00 (20060101) |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
202323092 |
|
Jul 2012 |
|
CN |
|
101880899 |
|
Sep 2012 |
|
CN |
|
0022113 |
|
Jul 1984 |
|
EP |
|
Other References
Hernandez et al., "Computer Aided Electroforming. Elecform3D",
2013, Procedia Engineering, vol. 63, pp. 532-539 (Year : 2013).
cited by examiner .
Chan et al., "Application of a hybrid case-based reasoning approach
in electroplating industry", Expert Systems with Applications, vol.
29, China, 2005,pp. 121-130. cited by applicant .
Castellano et al., "Computer Aided Electroforming for Rapid
Manufacturing Applications", ASME 2012 11th Biennial Conference on
Engineering Systems Design and Analysis, vol. 4, France, 2012,pp.
147-154. cited by applicant.
|
Primary Examiner: Ali; Mohammad
Assistant Examiner: Chang; Vincent W
Attorney, Agent or Firm: McGarry Bair PC
Claims
The invention claimed is:
1. A method of forming a component by an electroforming process
using an electroforming apparatus comprising an anode, a cathode
and an electrolyte comprising a metal salt, the method comprising:
receiving a set of training electroforming process parameters, the
set of training electroforming process parameters including one or
more of first coordinates of a shield, a first orientation state of
the shield, second coordinates of an auxiliary electrode, a second
orientation state of the auxiliary electrode, or a waveform
characteristic of an applied electric current; training, based on
at least one simulated electroforming process, a machine learning
algorithm based on at least a subset of the set of training
electroforming process parameters and wherein training the machine
learning algorithm comprises training the machine learning
algorithm based on the set of training electroforming process
parameters corresponding to the at least one simulated
electroforming process, validating the machine learning algorithm
based on a set of measured electroforming process parameters
corresponding to at least one actual electroforming process, and
storing the validated machine learning algorithm as the trained
machine algorithm; generating a set of updated operating
electroforming parameters from the trained machine algorithm; and
operating the electroforming apparatus based on the set of updated
operating electroforming parameters, wherein operating the
electroforming apparatus comprises: applying the applied electric
current between the anode and the cathode in a presence of the
electrolyte, and depositing a plurality of metal layers on a
surface of the cathode to form the component.
2. The method of claim 1, wherein the set of training
electroforming parameters comprises an initial set of operating
parameters of the electroforming apparatus and an initial thickness
variation across a surface of the component formed by the
electroforming process.
3. The method of claim 2, wherein the initial thickness variation
is calculated by applying the initial set of operating parameters
of the electroforming apparatus to a computational
electrodeposition model.
4. The method of claim 2, wherein the set of initial operating
parameters also comprises one or more of third coordinates of the
anode, a third orientation state of the anode, fourth coordinates
of the cathode, or a fourth orientation state of the cathode.
5. The method of claim 4, wherein the set of updated operating
parameters comprises one or more of first updated coordinates of
the anode, a first updated orientation state of the anode, second
updated coordinates of the cathode, a second updated orientation
state of the cathode, third updated coordinates of the shield, a
third updated orientation state of the shield, fourth updated
coordinates of the auxiliary electrode, a fourth updated
orientation state of the auxiliary electrode, or an updated
waveform characteristic of the applied electric current.
6. The method of claim 5, wherein operating the electroforming
apparatus comprises moving, using a robotic assembly, one or more
of the cathode, the anode, the auxiliary electrode, or the shield,
based on one or more of the first updated coordinates of the anode,
the first updated orientation state of the anode, the second
updated coordinates of the cathode, the second updated orientation
state of the cathode, the third updated coordinates of the shield,
the third updated orientation state of the shield, the fourth
updated coordinates of the auxiliary electrode, or the fourth
updated orientation state of the auxiliary electrode.
7. The method of claim 5, wherein operating the electroforming
apparatus comprises varying the applied electric current, using a
programmable power supply, based on the updated waveform
characteristic of the applied electric current.
8. The method of claim 1, wherein training the machine learning
algorithm comprises maintaining a root mean square error of a
thickness variation across a surface of the component, formed by
the electroforming process, in a range from about 1 micrometer to
about 200 micrometers.
9. The method of claim 1, wherein training the machine learning
algorithm comprises: training the machine learning algorithm based
on a first subset of the set of training electroforming process
parameters corresponding to the at least one simulated
electroforming process; validating the machine learning algorithm
based on a second sub-set of the set of training electroforming
process parameters corresponding to the at least one simulated
electroforming process; and storing the validated machine learning
algorithm as the trained machine algorithm.
10. A method of forming a component by an electroforming process
using an electroforming apparatus comprising an anode, a cathode
and an electrolyte comprising a metal salt, the method comprising:
receiving a set of training electroforming process parameters, the
set of training electroforming process parameters including one or
more of first coordinates of a shield, a first orientation state of
the shield, second coordinates of an auxiliary electrode, a second
orientation state of the auxiliary electrode, or a waveform
characteristic of an applied electric current; training, based on
at least one simulated electroforming process, a machine learning
algorithm based on at least a first subset of the set of training
electroforming process parameters; validating the machine learning
algorithm based on a second subset of the set of training
electroforming process parameters corresponding to the at least one
simulated electroforming process; generating a set of updated
operating electroforming parameters from at least one of the
training or the validating, wherein the set of updated operating
parameters comprises one or more of first updated coordinates of
the shield, a first updated orientation state of the shield, second
updated coordinates of the auxiliary electrode, a second updated
orientation state of the auxiliary electrode, or an updated
waveform characteristic of the applied electric current; and
operating an electroforming apparatus based on the set of updated
operating electroforming parameters, wherein operating the
electroforming apparatus comprises: moving, using a robotic
assembly, the anode, based on the set of updated operating
electroforming parameters, applying the applied electric current
between the anode and the cathode in a presence of the electrolyte,
after each movement of the anode, and depositing a plurality of
metal layers on a surface of the cathode to form the component.
11. A system for electroforming a component, the system comprising:
an electroforming apparatus comprising an anode, a cathode, and an
electrolyte comprising a metal salt; and a controller operatively
coupled to the electroforming apparatus, the controller comprising:
a memory storing one or more processor-executable routines and a
machine learning algorithm; and one or more processors to execute
the one or more processor-executable routines which, when executed,
cause acts to be performed comprising: receiving a set of training
electroforming process parameters, the set of training
electroforming process parameters including one or more of first
coordinates of a shield, a first orientation state of the shield,
second coordinates of an auxiliary electrode, a second orientation
state of the auxiliary electrode, or a waveform characteristic of
an applied electric current; training, based on at least one
simulated electroforming process, a machine learning algorithm
based on at least a subset of the set of training electroforming
process parameters; validating the machine learning algorithm based
on a set of measured electroforming process parameters
corresponding to at least one actual electroforming process;
generating a set of updated operating electroforming parameters
from at least one of the training or the validating; and operating
the electroforming apparatus based on the set of updated operating
electroforming parameters, wherein operating the electroforming
apparatus comprises: applying the applied electric current between
the anode and the cathode in a presence of the electrolyte, and
depositing a plurality of metal layers on a surface of the cathode
to form the component.
12. The system of claim 11, wherein the set of training
electroforming parameters comprises a set of initial operating
parameters of the electroforming apparatus and an initial thickness
variation across a surface of the component formed by the
electroforming process.
13. The system of claim 12, wherein the initial thickness variation
is calculated by applying the set of initial operating parameters
of the electroforming apparatus to a computational
electrodeposition model stored in the memory.
14. The system of claim 12, wherein the set of initial operating
parameters comprises one or more of third coordinates of the anode,
a third orientation state of the anode, fourth coordinates of the
cathode, or a fourth orientation state of the cathode.
15. The system of claim 14, wherein the set of updated operating
parameters comprises one or more of first updated coordinates of
the anode, a first updated orientation state of the anode, second
updated coordinates of the cathode, a second updated orientation
state of the cathode, third updated coordinates of the shield, a
third updated orientation state of the shield, fourth updated
coordinates of the auxiliary electrode, a fourth updated
orientation state of the auxiliary electrode, or an updated
waveform characteristic of the applied electric current.
16. The system of claim 15, further comprising a robotic assembly
configured to move one or more of the cathode, the anode, the
auxiliary electrode, or the shield, based on one or more of the
first updated coordinates of the anode, the first updated
orientation state of the anode, the second updated coordinates of
the cathode, the second updated orientation state of the cathode,
the third updated coordinates of the shield, the third updated
orientation state of the shield, the fourth updated coordinates of
the auxiliary electrode, or the fourth updated orientation state of
the auxiliary electrode.
17. The system of claim 15, further comprising a programmable power
supply configured to vary the applied electric current, based on
the updated waveform characteristic of the applied electric
current.
18. The system of claim 11, wherein training the machine learning
algorithm comprises maintaining a root mean square error of a
thickness variation across a surface of the component formed by the
electroforming process in a range from about 1 micron to about 200
micrometers.
Description
BACKGROUND
Embodiments of the disclosure generally relate to methods and
systems of forming electroformed components. More particularly,
embodiments of the disclosure relate to methods and systems of
forming electroformed components with complex geometry.
Electroforming is an additive manufacturing process where metal
components are formed through electrolytic reduction of metal ions
(atom by atom) on the surface of a mandrel (cathode).
Electroforming is used to manufacture products across a range of
industries including healthcare, electronics, and aerospace.
Electroforming manufacturing process can offer several advantages
including efficiency, precision, scalability, or
cost-effectiveness. However, electroforming of components having
complex shapes may pose technical challenges. For example,
components having significant curvature, tight corners, lack of
symmetry, recessed or even internal features, may pose difficulties
in setting up the process to yield consistent and controllable part
thickness.
Accordingly, there remains a need for improved methods of
manufacturing electroformed components, particularly for
electroforming of components with complex geometry.
BRIEF DESCRIPTION
In one aspect of the description, a method of forming a component
by an electroforming process using an electroforming apparatus is
presented. The electroforming apparatus includes an anode, a
cathode and an electrolyte including a metal salt. The method
includes receiving a set of training electroforming process
parameters; training a machine learning algorithm based on at least
a subset of the set of training electroforming process parameters;
generating a set of updated operating electroforming parameters
from the trained machine learning algorithm; and operating the
electroforming apparatus based on the set of updated operating
electroforming parameters. The step of operating the electroforming
apparatus includes applying an electric current between the anode
and the cathode in the presence of the electrolyte, and depositing
a plurality of metal layers on a cathode surface to form the
component.
In another aspect of the description, a method of forming a
component by an electroforming process using an electroforming
apparatus is presented. The electroforming apparatus includes an
anode, a cathode and an electrolyte including a metal salt. The
method includes receiving a set of training electroforming process
parameters including coordinates of the anode, orientation states
of the anode, or a combination thereof; training a machine learning
algorithm based on at least a subset of the set of training
electroforming process parameters; generating a set of updated
operating electroforming parameters from the trained machine
learning algorithm, wherein the set of updated operating parameters
includes updated coordinates of the anode, updated orientation
states of the anode, or a combination thereof; and operating an
electroforming apparatus based on the set of updated operating
electroforming parameters. The step of operating the electroforming
apparatus includes: moving, using a robotic assembly, the anode,
based on the set of updated operating electroforming parameters,
applying an electric current between the anode and the cathode in
the presence of the electrolyte, after each movement of anode, and
depositing a plurality of metal layers on a cathode surface to form
the component.
In yet another aspect of the description, a system for
electroforming a component, is also presented. The system includes
an electroforming apparatus and a controller operatively coupled to
the electroforming apparatus. The electroforming apparatus includes
an anode, a cathode and an electrolyte including a metal salt. The
controller includes a memory storing one or more
processor-executable routines and a machine learning algorithm. The
controller further includes one or more processors to execute the
one or more processor-executable routines which, when executed,
cause acts to be performed. The acts to be performed include:
receiving a set of training electroforming process parameters;
training a machine learning algorithm based on at least a subset of
the set of training electroforming process parameters; generating a
set of updated operating electroforming parameters from the trained
machine learning algorithm; and operating the electroforming
apparatus based on the set of updated operating electroforming
parameters. Further, operating the electroforming apparatus
includes: applying an electric current between the anode and the
cathode in the presence of the electrolyte, and depositing a
plurality of metal layers on a cathode surface to form the
component.
BRIEF DESCRIPTION OF THE DRAWINGS
These and other features, aspects, and advantages of the present
disclosure will become better understood when the following
detailed description is read with reference to the accompanying
drawings, wherein:
FIG. 1 is a schematic illustration of the electroforming process,
in accordance with some embodiments of the disclosure;
FIG. 2A shows a simulation of a baseline process for electroforming
a tubular junction using a cathode and a plurality of anodes;
FIG. 2B shows the thickness distribution for the tubular junction
formed using the simulation of FIG. 2A;
FIG. 3A shows a simulation of a baseline process for electroforming
a tubular junction using a cathode, a plurality of anodes, a
plurality of auxiliary anodes, a plurality of auxiliary cathodes,
and a plurality of shields;
FIG. 3B shows the thickness distribution for the tubular junction
formed using the simulation of FIG. 3A;
FIG. 4 is a schematic illustration of a system for electroforming a
component, in accordance with some embodiments of the
disclosure;
FIG. 5 is a flow chart illustrating an electroforming process, in
accordance with some embodiments of the disclosure;
FIG. 6 is a flow chart illustrating a step of training the machine
learning algorithm during the electroforming process, in accordance
with some embodiments of the disclosure;
FIG. 7 is a flow chart illustrating a step of training the machine
learning algorithm during the electroforming process, in accordance
with some embodiments of the disclosure;
FIG. 8 is a flow chart illustrating an electroforming process, in
accordance with some embodiments of the disclosure;
FIG. 9 shows a 2-D simulation of an electroforming process using a
contoured cathode and an oblong anode, in accordance with some
embodiments of the disclosure;
FIG. 10A shows the thickness of the layer deposited at time t=1 for
different points along the cathode surface, using the 2-D
simulation of FIG. 9, in accordance with some embodiments of the
disclosure;
FIG. 10B shows the reward values at time t=1 for different points
along the cathode surface, using the 2-D simulation of FIG. 9, in
accordance with some embodiments of the disclosure;
FIG. 11 shows the coordinates and orientation states of the oblong
anode used in a 2-D simulation, for every one second step-change,
in accordance with some embodiments of the disclosure;
FIG. 12A shows the anode kinematic pattern (degrees) for the
simulation of FIG. 11;
FIG. 12B shows the anode kinematic pattern (Y-coordinates) for the
simulation of FIG. 11;
FIG. 12C shows the anode kinematic pattern (X-coordinates) for the
simulation of FIG. 11; and
FIG. 13 shows the variation in thickness across different positions
along the cathode surface for a dynamic electrodeposition process
versus a static electroforming process, in accordance with some
embodiments of the disclosure.
DETAILED DESCRIPTION
In the following specification and the claims, which follow,
reference will be made to a number of terms, which shall be defined
to have the following meanings. The singular forms "a", "an" and
"the" include plural referents unless the context clearly dictates
otherwise. As used herein, the term "or" is not meant to be
exclusive and refers to at least one of the referenced components
being present and includes instances in which a combination of the
referenced components may be present, unless the context clearly
dictates otherwise.
Approximating language, as used herein throughout the specification
and claims, may be applied to modify any quantitative
representation that could permissibly vary without resulting in a
change in the basic function to which it is related. Accordingly, a
value solidified by a term or terms, such as "about", and
"substantially" is not to be limited to the precise value
specified. In some instances, the approximating language may
correspond to the precision of an instrument for measuring the
value. Similarly, "free" may be used in combination with a term,
and may include an insubstantial number, or trace amounts, while
still being considered free of the solidified term. Here and
throughout the specification and claims, range limitations may be
combined and/or interchanged, such ranges are identified and
include all the sub-ranges contained therein unless context or
language indicates otherwise.
As used herein, the terms "processor" and "computer" and related
terms, e.g., "controller" are not limited to just those integrated
circuits referred to in the art as a computer, but broadly refers
to a microcontroller, a microcomputer, a programmable logic
controller (PLC), an application specific integrated circuit, and
other programmable circuits, and these terms are used
interchangeably herein.
In the embodiments described herein, memory may include, but is not
limited to, a computer-readable medium, such as a random-access
memory (RAM), and a computer-readable non-volatile medium, such as
flash memory. Alternatively, a floppy disk, a compact disc--read
only memory (CD-ROM), a magneto-optical disk (MOD), flash drive,
and/or a digital versatile disc (DVD) may also be used. Also, in
the embodiments described herein, additional input channels may be,
but are not limited to, computer peripherals associated with an
operator interface such as a mouse and a keyboard. Alternatively,
other computer peripherals may also be used that may include, for
example, but not be limited to, a scanner. Furthermore, in the
exemplary embodiment, additional output channels may include, but
not be limited to, an operator interface monitor.
Further, as used herein, the term "non-transitory computer-readable
media" is intended to be representative of any tangible
computer-based device implemented in any method or technology for
short-term and long-term storage of information, such as,
computer-readable instructions, data structures, program modules
and sub-modules, or other data in any device. Therefore, the
methods described herein may be encoded as executable instructions
embodied in a tangible, non-transitory, computer readable medium,
including, without limitation, a storage device and/or a memory
device. Such instructions, when executed by a processor, cause the
processor to perform at least a portion of the methods described
herein.
Moreover, as used herein, the term "non-transitory
computer-readable media" includes all tangible, computer-readable
media, including, without limitation, non-transitory computer
storage devices, volatile and nonvolatile media, and removable and
non-removable media such as a firmware, physical and virtual
storage, a compact disc read only memory (CD-ROM), or a digital
versatile disc (DVD). The non-transitory computer storage devices
may also include digital source such as a network or the Internet,
as well as yet to be developed digital means, with the sole
exception being a transitory, propagating signal. Other
non-limiting examples of the memory include a dynamic random-access
memory (DRAM) device, a static random-access memory (SRAM) device,
and a flash memory.
As mentioned earlier, electroforming is an additive manufacturing
process where metallic components are grown on an appropriately
shaped cathode (usually referred to in the art as "mandrel")
through the electrochemical reduction of metal ions in a liquid
solution. FIG. 1 is a schematic illustration of an electroforming
process. As shown in FIG. 1, in the electroforming process, a
cathode 110 and an anode 120 are immersed in an electrolyte 130
solution and component thickness builds on the cathode surface 111
over time as current is passed between the electrodes 110, 120 (as
shown in FIG. 1). Once the desired component thickness is reached,
the cathode may be removed by mechanical, chemical, or thermal
treatment, yielding a free-standing metal component 140.
However, as noted earlier, electroforming of components having
complex shapes may pose challenges with respect to achieving
consistent and controllable wall thickness. Therefore, for complex
geometries, a typical electroforming process may involve a
combination of shields and/or auxiliary electrodes. The additional
shields and/or auxiliary electrodes may be placed in the
electroforming apparatus along with the anode and cathode, followed
by application of a specified current waveform over a fixed time.
Over time, the process parameters may be iterated over many
configurations, sometimes using computer modeling software, to set
up a new procedure for a specific component geometry. This modeled
and optimized procedure may specify the desired process parameters,
e.g., the size and shape of the anode(s), use and configuration of
auxiliary electrodes and/or shields, or precise management of
solution flow rates and direction. While this method can be
engineered to produce the desired component geometry, it is often a
time-consuming trial-and-error process that may not yield
components with an optimal thickness distribution.
By way of example, FIG. 2A shows a simulation (using COMSOL
Metaphysics.RTM. software) of a baseline process for electroforming
a tubular junction. The process of FIG. 2A is simulated using a
cathode 110 and a pair of simple flat sheet anodes 120. As shown in
FIG. 2B, thickness of the component 140 formed using the simulated
electroforming process is found to vary widely from about 30 mm in
the fillets, to over 100 mm at the tube edges. This is, at least in
part, due to the uneven current distribution and ionic
concentration gradients that are set up during the electroforming
process.
FIG. 3A shows a simulation result where a complex combination of
auxiliary anodes and cathodes, and shields are used to produce a
significantly more uniform component. FIG. 3A shows a simulation of
an electroforming setup include a cathode 110, anodes 120,
auxiliary cathodes 112, auxiliary anodes 122, and shields 124.
However, as shown in FIG. 3B, thickness variation was still
observed in the component 140, even after multiple iterations over
the complex simulation setup used here.
Embodiments of the present disclosure address the problems related
to electroforming of components with complex geometry by
integrating artificial intelligence in the electroforming process.
Embodiments of the present disclosure use trained machine learning
algorithms to optimize the electroforming of arbitrary, complex
shapes by adding several new degrees of freedom to the
electroforming process such as cathode and anode motion, custom
anode shape(s), varying current, and active control over shields or
auxiliary electrodes. Integration of the trained machine learning
algorithm with robotic manipulation enables reduction of the
process development time, and also enables manufacturing of more
complex components with minimal thickness variation, using
electroforming.
In some embodiments, a method of forming a component by an
electroforming process using an electroforming apparatus is
presented. The electroforming apparatus includes an anode, a
cathode and an electrolyte including a metal salt. The method
includes receiving a set of training electroforming process
parameters; training a machine learning algorithm based on at least
a subset of the set of training electroforming process parameters;
generating a set of updated operating electroforming parameters
from the trained machine learning algorithm; and operating the
electroforming apparatus based on the set of updated operating
electroforming parameters. Further, operating the electroforming
apparatus includes: applying an electric current between the anode
and the cathode in the presence of the electrolyte, and depositing
a plurality of metal layers on a cathode surface to form the
component.
In some embodiments, a system for electroforming a component, is
also presented. The system includes an electroforming apparatus and
a controller operatively coupled to the electroforming apparatus.
The electroforming apparatus includes an anode, a cathode and an
electrolyte including a metal salt. The controller includes memory
storing one or more processor-executable routines and a machine
learning algorithm. The controller further includes one or more
processors to execute the one or more processor-executable routines
which, when executed, cause acts to be performed. The acts to be
performed include: receiving a set of training electroforming
process parameters; training a machine learning algorithm based on
at least a subset of the set of training electroforming process
parameters; generating a set of updated operating electroforming
parameters from the trained machine learning algorithm; and
operating the electroforming apparatus based on the set of updated
operating electroforming parameters. Further, operating the
electroforming apparatus includes: applying an electric current
between the anode and the cathode in the presence of the
electrolyte, and depositing a plurality of metal layers on a
cathode surface to form the component.
The methods and systems of electroforming, in accordance with some
embodiments of the disclosure, are now described with reference to
FIGS. 4 and 5. FIG. 4 is an illustration of a system 200 of
electroforming a component and FIG. 5 depicts a flowchart
illustrating a method 10 of forming an electroformed component, for
example, using the system 200, in accordance with some embodiments
of the disclosure.
Referring now to FIG. 4, the system 200 includes an electroforming
apparatus 100, a controller 300, a robotic assembly 400 and a power
supply 500. In the embodiment illustrated in FIG. 4, the
electroforming apparatus 100 includes a cathode 110, an anode 120,
an electrolyte 130 disposed between the cathode 110 and the anode
120. As mentioned previously, the metal ions from the metal salt in
the electrolyte 130 are deposited on the cathode surface during the
electroforming process to form the component. In FIG. 4, the shape,
size and configuration of the cathode and the anode are shown for
illustration purposes only. The shape and size of the cathode will
depend on the geometry of the component to be manufactured using
the electroforming process. Once the desired component thickness is
reached, the cathode 110 may be removed by mechanical, chemical, or
thermal treatment, yielding a free-standing metal component. In one
example, the cathode 110 can be a low melting point material (i.e.
a "fusible alloy") which can be cast into the cathode shape and
subsequently melted out for re-use following electroforming. Other
cathode 110 options include conductive waxes and metallized plastic
which can be formed by injection molding, 3D printing, etc. In some
cases, a reusable cathode 110 may also be possible where component
geometry allows.
Further, in accordance with embodiments of the present disclosure,
the shape, the size, and the configuration (defined by the
coordinates and orientation states) of the anode 120 may be
determined and controlled using the machine learning algorithm, as
described in detail later. Further, in some embodiments, the
electroforming apparatus may include a plurality of anodes 120. In
such instances, the shape, the size, and the configuration of each
of the anode of the plurality of anodes 120 may be determined and
controlled using the machine learning algorithm.
In some embodiments, the electroforming apparatus may further
include one or more auxiliary cathode 112, one or more auxiliary
anode 122, or one or more shield 124, as illustrated in FIG. 3A.
The additional anodes, usually referred to as "auxiliary anodes" in
the art, are typically included in an electroforming process to add
additional local current density over a particular surface area of
the cathode. This is commonly needed over locations on the cathode
surface that would otherwise be too thin without the auxiliary
anode(s) in place. Similarly, auxiliary cathodes, sometimes
referred to as "thieves" in the art, are used to actively draw
excess current away from the main cathode (i.e. mandrel) so as to
avoid excessive buildup of metal thickness on the final part.
Auxiliary cathodes are often used near mandrel edges where part
thickness can build excessively. Along with or instead of auxiliary
anodes and cathodes, "shields" may also be used to block or channel
current between the anode and cathode to achieve a specified
deposition rate and component thickness. Shields are produced from
non-conductive materials and are placed in an electroforming
solution between the anode and cathode to shape the current density
distribution as required over a given mandrel surface.
The cathode 110, the anode 120, and the auxiliary electrodes 112,
122 may be collectively referred to as "electrodes", herein
throughout the specification. In FIG. 3A, the shape, size and
configuration of the auxiliary cathodes 112, the auxiliary anodes
122, and shields 124 are shown for illustration purposes only. The
shape, size and configuration of the auxiliary cathodes 112, the
auxiliary anodes 122, and shields 124 will depend on the geometry
of the component to be manufactured using the electroforming
process. Further, in accordance with embodiments of the present
disclosure, the shape, the size, and the configuration of the
auxiliary cathodes 112, the auxiliary anodes 122, and shields 124
may be determined and controlled using the machine learning
algorithm, as described in detail later. Furthermore, as shown in
FIG. 3A, the electroforming apparatus may include a plurality of
auxiliary cathodes 112, auxiliary anodes 122, and shields 124; the
shape, the size, and the configuration of each of the auxiliary
cathode 112, the auxiliary anode 122, and shields 124 may be
determined and controlled using the machine learning algorithm.
Referring again to FIG. 4, the electroforming apparatus 100 further
includes an electrolyte 130 disposed between the cathode 110 and
the anode 120. In some embodiments, the electrolyte 130 may be
further disposed between the cathode 110, the anode 120, the
auxiliary cathodes 112, the auxiliary anodes 122, and shields 124.
The electrolyte 130 includes a metal salt. In some embodiments, the
electrolyte includes a solution of the metal salt. Non-limiting
examples of suitable metal salts include chlorides, sulfates,
and/or sulfamates of nickel, copper, cobalt, or combinations
thereof. In certain embodiments, the metal salt includes a salt of
nickel. The electrolyte may also include other chemical additives
e.g., dispersants, surfactants and the like.
With continued reference to FIG. 4, in some embodiments, the
electroforming system 200 may further include a robotic assembly
400. In the embodiment shown in FIG. 4, the robotic assembly 400 is
depicted as mechanically coupled to the anode 120. In such
embodiments, the robotic assembly 400 may be configured to move the
anode 120 based on the configuration required for the
electroforming process. In some such embodiments, the robotic
assembly 400 may be configured to move the anode 120 based on a
signal received from the controller 300. In some embodiments, the
robotic assembly 400 may be instead or additionally configured to
move one or more of the cathode 110, the auxiliary cathodes 112,
the auxiliary anodes 122, and the shields 124. In some embodiments,
the electroforming system 200 may include a plurality of such
robotic assemblies 400.
As shown in FIG. 4, the electroforming system 200 further includes
a power supply 500 communicatively linked to the controller 300.
The power supply 500 applies a current between the cathode 110 and
the anode 120, thereby depositing a plurality of metal layers on
the cathode surface 111 to form the component, by directing metal
ions from the metal salt onto the cathode surface 111. In some
embodiments, the power supply 500 may be a programmable power
supply 500 that is communicatively linked to the controller 300.
The programmable power supply 500 may be configured to vary the
current waveform that is applied between the cathode 110 and the
anode 120, based on a signal received from the controller 300. In
some embodiments, the programmable power supply 500 may be
configured to vary the current waveform that is applied between the
cathode 110, the anode 120, the auxiliary cathode 112, and the
auxiliary anodes 122, shields, based on a signal received from the
controller 300.
As noted earlier, the electroforming system further includes a
controller 300, as shown in FIG. 4. The controller 300 includes one
or more processors, such as, a processor 310. The processor 310 may
include a specially programmed general-purpose computer, a
microprocessor, a digital signal processor, and a microcontroller.
Examples of the processor 310 include, but are not limited to, a
reduced instruction set computing (RISC) architecture type
processor or a complex instruction set computing (CISC)
architecture type processor. Further, the processor 310 may be a
single-core processor or a multi-core processor. The processor 310
may also include, or, has electrically coupled thereto, one or more
input/output ports.
The controller 300 further includes a memory 320 accessible by the
processor 310. In some embodiments, the memory 320 may be
integrated into the processor 310. In some other embodiments, the
memory 320 may be external to the processor 310 and electrically
coupled to the processor 310, as depicted in FIG. 4. The memory 320
may be a non-transitory computer-readable media. The non-transitory
computer-readable media may include tangible, computer-readable
media, including, without limitation, non-transitory computer
storage devices.
The memory 320 stores processor-executable routines that are
executable by the processor 310. The processor-executable routines,
when executed by the processor 310, cause acts to be performed that
contribute to methods described below as well as other variants
that are anticipated, but not specifically listed. In a
non-limiting example, processor-executable routines may be
implemented in a variety of programming languages, including but
not limited to C, C++, or Java. In some embodiments, by executing
one or more of the processor-executable routines, the processor 310
may aid in manufacturing a component using an electroforming
process.
The processor-executable routines, when executed by the processor
310 cause acts to be performed. The acts to be performed include
steps illustrated in flowcharts of FIG. 5, in accordance with some
embodiments of the disclosure. As illustrated in FIG. 5, a method
10 of forming a component using an electroforming process includes
steps 20-50. The processor-executable routines, when executed by
the processor 310, may cause the processor 310 to perform acts
indicated by the steps 20-50 of the method 10.
Referring now to FIGS. 3-5, the method 10 includes receiving a set
of training electroforming process parameters at step 20. In some
embodiments, the set of training electroforming process parameters
may be stored as a database in the memory 320 of the controller
300. In such instances, the step 20 of receiving may include
retrieving the set of training electroforming parameters from the
memory 320 via the one or more processor 310. In some other
embodiments, the set of training electroforming process parameters
may be provided as an input by a user to the one or more processor
310 of the controller 300, e.g., via a user input interface (not
shown).
In some embodiments, the set of training electroforming parameters
include an initial set of operating parameters of the
electroforming apparatus 100 and an initial thickness variation
across a surface of the component formed by the electroforming
process. The initial operating parameters may include one or more
of coordinates of the anode 120, orientation states of the anode
120, coordinates of the cathode 110, orientation states of the
cathode 110, coordinates of the shield 124, orientation states of
the shield 124, coordinates of an auxiliary electrode 112, 122,
orientation states of an auxiliary electrode 112, 122, or a
waveform characteristic of the applied electric current. The term
"coordinates" as used herein refers to the location of the
electrode, e.g., the anode in the xyz plane and can be represented
by x.sub.t, y.sub.t, and z.sub.t, wherein "t" is the time at which
the coordinates are determined. The term "orientation states" as
used herein refers to the angles that the electrode, e.g., the
anode makes in the xyz plane and can be represented by
.theta..sub.xy, .theta..sub.yz, and .theta..sub.xz at time "t".
Similarly, the coordinates and the orientation states of the
cathode 110, the auxiliary electrodes 112, 122 and shields 124 may
be represented at a time "t". In some embodiments, the initial
operating parameters of the electroforming apparatus 100 may be
based on historical data. In some embodiments, the historical data
may be based on actual operating parameters employed during an
electroforming process. In some other embodiments, the historical
data may be based on an estimate of the operating parameters by the
operator based on previous experience.
The term "initial thickness variation" as used herein refers to the
variation in thickness across a surface of the component to be
formed before the application of the machine learning algorithm. In
some embodiments, the initial thickness variation may be based on
historical data. In some embodiments, the initial thickness
variation may be calculated by applying the initial set of
operating parameters of the electroforming apparatus 100 to a
computational electrodeposition model. Non-limiting examples of
suitable computational electrodeposition models include solutions
of Tafel, Butler-Volmer, and/or Nernst-Plank equations using, for
example, a finite element numerical approach.
Referring again to FIGS. 4 and 5, the method further includes, at
step 30, training a machine learning algorithm based on at least a
subset of the set of training electroforming process parameters. As
noted earlier, the machine learning algorithm may be stored in the
memory 320 of the controller 300. The machine learning algorithm
may employ supervised learning, unsupervised learning, or
reinforcement learning. In some embodiments, a suitable machine
leaning algorithm in accordance with embodiments of the disclosure
is based on reinforcement learning. Non-limiting example of a
suitable reinforcement learning-based algorithms include Q-learning
algorithm. In some embodiments, the machine learning algorithm is
based on deep learning such as a deep neural network. In some
embodiments, the machine learning algorithm is a deep reinforcement
learning algorithm, e.g., a deep Q-learning algorithm.
In certain embodiments the machine learning algorithm is a
Q-learning algorithm. Q-learning is a reinforcement learning
technique used in machine learning. Q-learning is a model-free
learning environment that can be used in situations where the agent
initially knows only the possible states and actions but doesn't
know the state-transition and reward probability functions. In
Q-learning the agent improves its behavior (online) through
learning from the history of interactions with the environment.
Q-learning involves an agent, a set of states ("S") and a set of
actions ("A") per state. The goal of Q-learning is to learn a
policy, which tells an agent what action to take under what
circumstances. By performing an action "a" within the set of
actions A (a .epsilon. A), the agent transitions from state to
state. Executing an action in a specific state provides the agent
with a reward (a numerical score). The goal of the agent is to
learn to select actions to maximize its total (future) reward. In
particular, in Q-learning, the reinforcement learner tries to infer
an action-value function, i.e., a function which predicts the value
(in terms of the reward that will be achieved) of each of the many
actions an agent could take. Thus, if the approximation is good,
the agent can choose the best action. It does this by updating the
approximation of the Q-function after taking an action and
observing the reward by adding the old estimate of the reward for
the chosen state-action pair with the discounted predicted future
reward.
In some embodiments of the disclosure, the reward as used in the
Q-learning algorithm is defined as the root mean square error
(RMSE) of a thickness variation across a surface of the component,
formed by the electroforming process 10. In such instances, high
RMSE would imply lower rewards. The term "thickness variation" as
used herein refers to the difference (plus or minus) in the
thickness between a pre-determined target thickness and the
thickness obtained after the electrodeposition process (e.g., after
the simulated electrodeposition process). In some embodiments, the
target thickness may be substantially uniform across the component.
In some other embodiments, the component may be defined by a
pre-determined target thickness distribution, i.e., the component
may include thicker and thinner portions. In some such instances,
the component may have a substantially non-uniform target thickness
and may be characterized by a target thickness profile. In such
embodiments, the reward function (e.g., defined by RMSE) may be
based on the deviation from the desired thickness profile.
Embodiments of the disclosure, as discussed herein, allow for
training of the machine learning algorithm such that a component
with a target distributed thickness profile may be
manufactured.
Further, as noted earlier, the goal is to maximize rewards and thus
lower RMSE. In accordance with some embodiments of the disclosure,
training the machine learning algorithm includes maintaining a root
mean square error (RMSE) of a thickness variation across a surface
of the component, formed by the electroforming process, in a range
from about 1 micrometer to about 200 micrometers. In accordance
with some embodiments of the disclosure, training the machine
learning algorithm includes maintaining a root mean square error
(RMSE) of a thickness variation in a range from about 1 micrometer
to about 50 micrometers. In accordance with some embodiments of the
disclosure, training the machine learning algorithm includes
maintaining a root mean square error (RMSE) of a thickness
variation in a range from about 1 micrometer to about 10
micrometers. Further, in some embodiments, depending on the end
application of the electroformed component, the RMSE of a thickness
variation may be less than 1 micrometer or greater than 200
micrometers. In some such embodiments, the RMSE of a thickness
variation may be in a range from about 1 nm to about 1 micrometer.
Further, in some such embodiments, the RMSE of a thickness
variation may be in a range from about 200 micrometers to about
10000 micrometers.
In some embodiments, the trained machine learning algorithm may be
further validated and stored in the memory 320 of the controller
300. Referring now to FIG. 6, in some embodiments, the step 30 of
training the machine learning algorithm includes, at sub-step 31,
training the machine learning algorithm based on the set of
training electroforming process parameters. As noted earlier, in
some embodiments, a computer simulation may be used to train the
machine learning algorithm. In such instances, the entire set of
training electroforming parameters may used to train the machine
learning algorithm. The step 30 further includes, at sub-step 32,
validating the machine learning algorithm based on a set of
measured electroforming process parameters. The set of measured
electroforming parameters may be obtained in real-time via
appropriate sensors, or, alternately may be based on historical
data that may be, e.g., stored in the memory of the controller 300.
The step 30 furthermore includes, at sub-step 33, storing the
validated machine learning algorithm as a trained machine
algorithm.
In some other embodiments, as shown in FIG. 7, the step 30 of
training the machine learning algorithm includes, at sub-step 34,
training the machine learning algorithm based on a first subset of
the set of training electroforming process parameters. As noted
earlier, in some embodiments, a computer simulation may be used to
train the machine learning algorithm. The step 30 further includes,
at sub-step 35, validating the machine learning algorithm based on
a second sub-set of the set of training electroforming process
parameters. The step 30 furthermore includes, at sub-step 36,
storing the validated machine learning algorithm as a trained
machine algorithm.
Referring back to FIGS. 3-5, the method 10 further includes, at
step 40, generating a set of updated operating electroforming
parameters from the trained machine learning algorithm. As
described herein earlier, the trained machine learning algorithm
may be validated and stored in the memory 320 of the controller
300. In some embodiments, the set of updated operating parameters
includes one or more of updated coordinates of the cathode 110,
updated orientation states of the cathode 110, updated coordinates
of the anode 120, updated orientation states of the anode 120,
updated coordinates of the shield 124, updated orientation states
of the shields 124, updated coordinates of the auxiliary electrodes
112, 122, updated orientation states of the auxiliary electrode
112, 122, or an updated waveform characteristic of the applied
electric current.
The method further includes, at step 50, operating the
electroforming apparatus 100 based on the set of updated operating
electroforming parameters. In some embodiments, operating the
electroforming apparatus 100 includes moving, using a robotic
assembly 400, one or more of the cathode 110, the anode 120, the
auxiliary electrodes 112, 122, or the shields 124, based on one or
more of the updated coordinates of the cathode 110, updated
orientation states of the cathode 110, updated coordinates of the
anode 120, updated orientation states of the anode 120, updated
coordinates of the shields 124, updated orientation states of the
shields 124, updated coordinates of the auxiliary electrodes 112,
122, or updated orientation states of the auxiliary electrodes 112,
122. As mentioned earlier, FIG. 4 shows a single robotic assembly
400 mechanically coupled to the anode 120 and an end-effector of
the robotic assembly is configured to move the anode 120. However,
the robotic assembly may be further coupled to one or more of the
cathode 110, the auxiliary electrodes 112, 122, or the shields 124.
Further, the method 10 may include moving the anode 120, the
cathode 110, the auxiliary electrodes 112, 122, or the shields 124,
using a plurality of robotic assemblies 400.
In some embodiments, after step 40, the controller 300 may send a
signal to the robotic assembly 400 to move one or more of the anode
120, the cathode 110, the auxiliary electrodes 112, 122, or the
shields 124. The signal from the controller 300 to the robotic
assembly 400 may further include details of the type (e.g., upward,
downward, sideward, angular) and degree of movement (e.g.,
displacement distance, angle of rotation) desired based on the
updated coordinates and/or the orientation states. In some other
embodiments, if after the step 40, there is no change in the
updated coordinates and/or the orientations states of the
electrodes in the electroforming apparatus 100, the controller 300
may either send a signal indicating no change, or, alternatively,
no signal may be sent from the controller 300 to the robotic
assembly 400.
In some embodiments, operating the electroforming apparatus 100
includes varying the applied electric current, using the
programmable power supply 500, based on the updated waveform
characteristic of the applied electric current. In some
embodiments, after step 40, the controller 300 may send a signal to
the programmable power supply 500 to vary the current waveform. The
signal from the controller 300 to the programmable power supply 500
may further include details of the current waveform desired based
on the updated waveform. In some other embodiments, if after the
step 40, there is no change in the updated current waveform, the
controller 300 may either send a signal indicating no change, or,
alternatively, no signal may be sent from the controller 300 to the
programmable power supply 500.
In some embodiments, after step 40, the controller may send a
signal to both the robotic assembly 400 and the programmable power
supply 500, based on the updated coordinates of the electrodes or
the shields, the updated orientations states of the electrodes or
the shields, and the updated waveform characteristic of the applied
electric current. In such instances, the method may include moving
one or more the electrodes or the shields, as well as changing the
waveform characteristic of the applied electric current. In
accordance with some embodiments of the disclosure, the movement of
the electrodes or the shields and/or the application of updated
current waveform may be executed in real-time.
Referring again to FIGS. 3-5, the step 50 of method 10 further
includes, at sub-step 51, applying an electric current between the
anode 120 and the cathode 110 in the presence of the electrolyte.
In some embodiments the electric current applied between the anode
and the cathode may be based on the updated waveform characteristic
as determined by the trained machine learning algorithm. Further,
in some embodiments, one or both of the anode 120 and the cathode
110 may be moved to the updated coordinates and/or orientations
states before the application of the electric current. In
embodiments including shields 124 and/or auxiliary electrodes
112,122, one or more these may also be moved before the application
of the electric current. In some embodiments, the method 10 may
include applying the electric current between the anode 120 and the
cathode 110, after each movement of the anode 120, the cathode 110,
the shields 124 and/or the auxiliary electrodes 112, 122. The step
50 further includes, at sub-step 52, depositing a plurality of
metal layers on a cathode surface 111 to form the component.
Without being bound by any theory, it is believed that the method
and systems employing trained machine learning algorithms, in
accordance with embodiments of the disclosure, enable manufacture
of components with complex geometries such that the components have
a minimal RMSE of thickness variation across the surface of the
components. As noted earlier, the component being manufactured
using the embodiments described herein may have a target thickness
profile that may be characterized by a substantially uniform
thickness across the component or a distributed thickness profile
across the component.
Referring now to FIGS. 4 and 8, a method 60 of forming a component
by an electroforming process, in accordance with one embodiment of
the disclosure, is presented. The method 60 includes, at step 61,
receiving a set of training electroforming process parameters
including coordinates of the anode 120, orientation states of the
anode 120, or a combination thereof. The method 60 further
includes, at step 62, training a machine learning algorithm based
on at least a subset of the set of training electroforming process
parameters. The method 60 further includes, at step 63, generating
a set of updated operating electroforming parameters from the
trained machine learning algorithm, wherein the set of updated
operating parameters includes updated coordinates of the anode 120,
updated orientation states of the anode 120, or a combination
thereof. The method 60 furthermore includes, at step 64, operating
an electroforming apparatus 100, based on the set of updated
operating electroforming parameters. The step 64 of operating the
electroforming apparatus includes, at sub-step 65, moving, using a
robotic assembly 400, the anode 120, based on the set of updated
operating electroforming parameters. The step 64, further includes,
at sub-step 66, applying an electric current between the anode 120
and the cathode 110 in the presence of the electrolyte 130, after
each movement of the anode 120. The step 64, further includes, at
sub-step 67, depositing a plurality of metal layers on a cathode
surface 111 to form the component, by directing metal ions from the
metal salt on the cathode surface 111. In one example, a custom
shaped anode 120 could be connected as the end effector of the
robotic assembly 400. During the electroforming process, the anode
120 may be moved, based on the updated coordinates and/or
orientation states from the trained machine learning algorithm, to
maintain near-constant component thickness over the entire cathode
surface 111.
Referring now to FIG. 9, a 2-D simulation of an electroforming
process using a cathode 110 and an anode 120, in accordance with
embodiments of the disclosure, is represented. The cathode 110 has
a curved contour, with a surface defined by 111, in the simulation
as a representation of a complex geometry of the electroformed
component. Further, in FIG. 9, the anode 120 is represented as
having an oblong shape. At time "t" during the electroforming
process, the anode 120 is characterized by a set of states
"S.sub.t" (1200 in FIG. 9). This set of states S.sub.t include the
coordinates of the anode at time "t": X.sub.t, Y.sub.t and Z.sub.t
as well as orientation states at time "t":.theta..sub.xy,
.theta..sub.yz, and .theta..sub.xz. The set of states at time "t"
is provided as an input to a machine learning algorithm, e.g., a
Q-learning algorithm whose goal is to find an optimal policy that
maximizes rewards. As noted earlier, in some embodiments, the
reward function is defined as RMSE, and low RMSE represents high
rewards.
In some embodiments, a Bellman equation may be used to optimize the
Q-function as shown by equation I:
Q*(s.sub.ta.sub.t)=E.sub.s.sub.t+1[r.sub.t+.gamma.max.sub.a.sub.t+1Q*(s.s-
ub.t+1a.sub.t+1)] (I) wherein, Q* is optimized Q function,
Es.sub.t+1 is the learning rate, g is the discount factor and is in
a range from 0 to 1, r.sub.t is the reward at time "t", s.sub.t is
the set of states at time "t", a.sub.t is the set of actions at
time "t", s.sub.t+1 is the set of states at time "t+1", and
a.sub.t+1 is the set of actions at time "t+1". Learning rate is
used to estimate the confidence level (in the range from 0 to 1) of
new value at each trial. A learning estimate of 1 means that all
new values from the reward network are used for the iteration.
Based on the Bellman equation, the agent learns the optimal
position and orientation to deposit uniform thickness. Further, by
maximizing the Q function, updated coordinates and orientation
states are derived, which translate into updated displacement and
orientation actions.
In order to learn from the past experience, epsilon
(.epsilon.)-greedy approach is used to choose updated states and
actions. In the .epsilon.-greedy approach at each step, with small
probability .epsilon., the agent picks a random action (i.e.,
explores) or with probability (1-.epsilon.) the agent selects an
action according to the current estimate of Q-values. In some
embodiments, the .epsilon. value is 0.5. Initially, the agent
randomly picks up states and actions. Following more iterations,
the agent starts to choose the maximal value from the Q matrix
(learning from good experiences). .epsilon. value can be decreased
overtime as the agent becomes more confident with its estimate of
Q-values.
In some embodiments, the reward function is calculated based on
equations (II)-(IV)
.times..function..function..times..times..function.>.times..times..tim-
es..times..function.< ##EQU00001## wherein RMSD=root mean square
deviation or root mean square error, y.sub.p=thickness at positions
"p", along the mandrel (cathode) surface, w=weight value, S=total
number of positions measured along the mandrel surface,
std=standard deviation of the thickness. N is the total grid
points/samples.
Referring now to FIGS. 10A and 10B, the thickness of the layer
deposited and reward values for different points along the mandrel
(cathode) surface, using the simulation of FIG. 9, at time t=1 are
shown. As shown in FIG. 10B, action "30" corresponds to the maximum
reward function at time t=1. After the completion of the protocol
at time "t=1", the initial condition at t+1 is replaced with the
deposited surface from the previous time step. In some embodiments,
each surface condition is unique and the agent cannot learn from
the previous time step. The initial states (S) and actions (A) at
each time step are set to be randomly chosen as initial guess for Q
learning and the protocols are repeated until the end of the
duration.
After policy validation, an updated policy from the machine
learning algorithm is generated which includes a set of Actions
A.sub.t', at time t'. The Actions A.sub.t', are based on the
updated coordinates and orientation states X.sub.t', Y.sub.t',
Z.sub.t', .theta..sub.xy', .theta..sub.yz', and .theta..sub.xz'. In
FIG. 9, the set of Actions are represented by the arrows 1201 and
curved arrow 1202. As shown in FIG. 9, based on the set of Actions
generated from the machine learning algorithm, the new set of
actions 1201 may represent upward motion, downward motion, motion
towards the left, or motion towards the right. Further, the action
1202 may represent rotation of the anode 120 based on the updated
orientation states of the anode 120. Thus, by employing the trained
Q-learning-based algorithm, at least six degrees of freedom for the
anode 120 (X, Y, Z, .theta..sub.xy, .theta..sub.yz, and
.theta..sub.xz) can be varied to maximize the rewards and minimize
RMSE. This may be further complemented with addition degrees of
freedom for the cathode 110, the auxiliary electrodes 112, 122, and
the shields 124, as described earlier.
Thus, embodiments of the disclosure use artificial intelligence
methods coupled with process modeling and robotics to produce
complex electroformed shapes in a more efficient way. Computational
electrodeposition models are used to teach an appropriate machine
learning algorithm where additional degrees-of-freedom are allowed
in the electroforming process. The additional degrees of freedom
can include real-time multi-axis motion of the cathode and/or
anode, custom shaped anodes, moveable shields and/or auxiliary
electrodes, and/or the option of varying the applied current
waveform over time. After model training, and once an acceptable
process is simulated, the prescribed motion(s) may be applied
during the electroforming process using robotic controls and/or the
current waveform may be applied using a programmable power supply.
It is expected that this approach will result in better quality
components with shorter development time.
This is in contrast to conventional electroforming processes that
are for the most part, static. That is, once a configuration is
set, the anode, cathode, and any auxiliary features are placed in a
tank and power is applied to build the component(s). There is no
active feedback or ability to correct uneven deposition during the
duration of the process. In accordance with embodiments of the
disclosure, electrodeposition models and artificial intelligence
methods are used to optimize the additional degrees-of-freedom to
produce high quality components (e.g., with minimal thickness
variation) having complex geometry. Further, the integration of
machine learning with robotic manipulation may also reduce the
process development time.
As noted earlier, the methods and systems in accordance with
embodiments of the disclosure may advantageously allow for
fabrication of electroformed components having a complex geometry
with a substantially uniform thickness. In some embodiments, the
electroforming methods and systems, as described herein, may be
suitable for manufacturing components with dimensions greater than
100 microns. In some other embodiments, the electroforming methods
and systems, as described herein, may suitable for manufacturing of
components with dimensions on the orders of a few microns, or even
in the range of nanometers. In some embodiments, the electroformed
component is a component of an aircraft engine, a gas turbine, or a
marine engine. In certain embodiments, the electroformed component
includes aircraft engine conveyance components, tubings, ducting,
seals, vanes, airfoils, struts, liners, cases, flow-path
structures, leading edges, brackets, flanges, or housings. In some
embodiments, the electroformed component may be a component of an
industry where small-scale parts are desired with micro-precision,
e.g., optical components, surgical instruments, medical
instruments, scientific instrumentation, microelectronics,
microfluidic devices, microelectronic mechanical systems (MEMS),
sensors, actuators, nanostructures, and the like.
EXAMPLES
The examples that follow are merely illustrative and should not be
construed to be any sort of limitation on the scope of the claimed
invention.
A 2-D simulation of an electroforming process using an
oblong-shaped anode 120 and a cathode 110 with a curvature (defined
by surface 111) is shown in FIG. 11. The simulation was setup with
1000 iterations with five episodes of Q learning process at each
time step. The initial epsilon rate was set to 0.5 and the decay
factor per iteration was set to 0.98. The learning rate for this
case was setup to be 0.99. The discount rate was set to 0.9 for
determining the importance of future state. The Q table determined
by resolution of states and actions was a matrix of 77 by 72. An
electroforming deposition process for a time duration of 10
seconds, by changing the coordinates and orientation states, of the
anode was simulated. FIG. 11 shows the coordinates and orientation
states of the oblong anode 120 (represented by solid line), used in
the simulation, after each stage of movement. The arrows indicate
the direction of movement of the anode. FIGS. 12A-12C further show
the anode kinematic pattern for the 10 seconds simulation. As
described in detail earlier, the coordinate and orientation states
as obtained from FIGS. 12A-12C can be used for manipulating the
trajectory of the robotic assembly
FIG. 13 shows the variation in thickness across different positions
along the mandrel (cathode) surface for a dynamic electroforming
process versus a static electroforming process. The simulation for
the dynamic process (e.g., a process that involves movement of the
anode 120) was conducted for 10 seconds, as described earlier, with
respect to FIGS. 11 and 12A-12C. The simulation for the static
process was also conducted for 10 seconds, however, in the static
process the anode was kept in a stationary position throughout the
simulation. As shown in FIG. 13, the dynamic electroforming process
showed significantly lower variation in thickness across the
surface as compared to a static electroforming process. In the
simulation set-up described herein, at least five times reduction
in surface thickness variation was obtained by using the dynamic
electroforming process, in accordance with the embodiments
described herein.
The appended claims are intended to claim the invention as broadly
as it has been conceived and the examples herein presented are
illustrative of selected embodiments from a manifold of all
possible embodiments. Accordingly, it is the Applicants' intention
that the appended claims are not to be limited by the choice of
examples utilized to illustrate features of the present disclosure.
As used in the claims, the word "comprises", and its grammatical
variants logically also subtend and include phrases of varying and
differing extent such as for example, but not limited thereto,
"consisting essentially of" and "consisting of" Where necessary,
ranges have been supplied; those ranges are inclusive of all
sub-ranges there between. It is to be expected that variations in
these ranges will suggest themselves to a practitioner having
ordinary skill in the art and where not already dedicated to the
public, those variations should where possible be construed to be
covered by the appended claims. It is also anticipated that
advances in science and technology will make equivalents and
substitutions possible that are not now contemplated by reason of
the imprecision of language and these variations should also be
construed where possible to be covered by the appended claims.
* * * * *